Table 5.
Results of independent sample t-test between throwaway mental health posts (MH) and throwaway control posts.
| Category | MH | Control | t-stat | p |
|---|---|---|---|---|
| Affective Attributes | ||||
| negative emotion | 0.0413 | 0.0239 | 9.476 | *** |
| anger | 0.0120 | 0.0077 | 4.415 | *** |
| anxiety | 0.0081 | 0.0055 | 4.118 | *** |
| sad | 0.0121 | 0.0037 | 6.289 | *** |
| Cognitive Attributes | ||||
| Cognition | ||||
| negation | 0.0309 | 0.0220 | 8.387 | *** |
| certainty | 0.0171 | 0.0131 | 4.800 | *** |
| death | 0.0043 | 0.0011 | 5.838 | *** |
| Perception | ||||
| see | 0.0048 | 0.0080 | −6.362 | *** |
| feel | 0.0107 | 0.0078 | 4.263 | *** |
| Linguistic Style Attributes | ||||
| Lexical Density | ||||
| nouns | 0.1777 | 0.2146 | −8.808 | *** |
| verbs | 0.1899 | 0.1699 | 4.679 | *** |
| adverbs | 0.0658 | 0.0566 | 3.264 | ** |
| Temporal References | ||||
| present tense | 0.1269 | 0.1033 | 8.293 | *** |
| past tense | 0.0391 | 0.0475 | −3.894 | *** |
| Social/Personal Concerns | ||||
| friends | 0.0018 | 0.0027 | −3.113 | ** |
| social | 0.0839 | 0.1241 | −8.200 | *** |
| work | 0.0183 | 0.0272 | −3.232 | ** |
| health | 0.0138 | 0.0102 | 3.826 | ** |
| humans | 0.0090 | 0.0151 | −4.438 | *** |
| money | 0.0039 | 0.0071 | −5.437 | *** |
| sexual | 0.0049 | 0.0073 | −4.203 | *** |
| Interpersonal Awareness | ||||
| 1st person singular | 0.1180 | 0.0802 | 15.910 | *** |
| 1st person plural | 0.0054 | 0.0112 | −4.429 | *** |
| 2nd person | 0.0045 | 0.0121 | −8.456 | *** |
| 3rd person | 0.0195 | 0.0295 | −6.260 | *** |
| Social Attributes | ||||
| number of comments | 5.2614 | 30.5753 | −14.350 | *** |
| vote difference | 6.0733 | 43.0134 | −8.968 | *** |
| median comment length | 78.5943 | 38.8837 | 7.112 | *** |
We report results at two different α levels: .01, .001. The significance values p above are the adjusted p's after adopting the Holm-Bonferroni method [17] which counteracts the problem of multiple comparisons here (total number of measures m = 52). This method is intended to control the Familywise error rate and offers a simple test uniformly more powerful than the Bonferroni correction [17].